作者：Hassan Baig

手机游戏已经成为一种全球性的巨大商机，在2012年的市场规模达90亿美元，在未来数年其产值还将进一步增长。预计到2015年全球智能手机将双倍增长，游戏在所有手机应用收益中也将占比高达66%。

游戏开发的前景依然光明，但这只适用于能够推出具有盈利性产品的团队。对于这种盈利性的追求已经形成了一门精确的科学，监测分析和持续的A/B测试也因此成为游戏开发的常态。事实上，Zynga这家推广了数据分析行为的游戏公司，通常被归为大型数据公司。

由此可见，参数已经成为游戏开发者用来实现其追求结果的“杠杆”。有些杠杆具有广泛的活动范围，而有些则具有局限性。游戏开发者的任务实际上是，以最少的成本找到能够产生最佳商业结果的完美杠杆组合。

revenue(from huanqiu.com)

至于创意、创新和趣味则要受到这种以分析为主的开发方法——它们有助于优化分析结果，才会有发挥空间。这正是当代游戏开发背后的基本原理。

为了方便各位进一步理解游戏数据分析的原理，我设置了一个简单的数学模拟法，通过假设性的留存率和病毒配置来对比游戏表现。

假设有家游戏工作室旗下有6款游戏：

gamer-metrics(from gamasutra)

这里我们以游戏1为基准，其他游戏都有一个参数与之不同。例如，游戏2在平均玩家生命周期上有别于游戏1（游戏邦注：其他各项参与均与之相同）。我已经使用平均为1.3美元的CPA来计算游戏各自的广告投入。现在我们来看看这种模拟方法的广泛结论：

1）游戏的玩家平均生命周期越长，其DAU数量就越高。由于DAU是衡量玩家粘性的近似测度，而粘性又直接与游戏收益挂钩，因此平均玩家生命周期对游戏潜在收益具有明显的影响。

在表格中的游戏1与游戏2的“相当收益”条目中，我们可以注意到游戏2有更长的玩家生命周期，相对于游戏1，它在DAU和收益数据上也更占优势。

2）游戏第二天留存率越高，DAU数量就越可观。DAU直接与收益相关，因此可以推测第二天留存率也会对游戏收益潜力产生明显影响。多数游戏公司在上线早期都会使用A/B测试来优化游戏留存率。此外，由于第二天留存率通常无法优化到2位数，所以拥有较低留存率的游戏很快就会被剔除。对比游戏1和游戏3就可以看到，后者的第二天留存率更高，所以DAU数量也更乐观，其收益也更为显著。

3）巨大的广告预算并不能提升游戏盈利性。也就是说，如果游戏在特定玩家群体中的收益不甚理想，收买更多同一群体的玩家也无济于事。所有，有许多游戏公司会在投入大量营销预算之前优化游戏参数，这也正是有些游戏尚未问世就被取消的原因。

4）游戏病毒性越强，盈利性越高。更强的病毒性确保游戏免费获取用户，这也就可以最小化一个关键成本因素——每用户获取成本。开发者如果拥有一个铁杆玩家网络也可以实现类似的效果，以较低的成本交叉推广自己的游戏。病毒性还可以实现玩家网络的自我扩展。

总之，获得免费用户对任何一家游戏公司来说都很重要，所以Mark Pincus才会在最近的财报中一再强调要把投入和利用Zynga玩家网络作为公司未来战略的基石。

运算达人能够一眼看到游戏1和游戏5之间的区别，分辨出这两款游戏的eCPA差异。

5）每用户收益越高，游戏利润就越可观。这是一个很显然的结论，但其中蕴含的深意却值得进一步探究。这正是游戏公司追求长期/多次游戏会话，以及43岁左右家庭主妇或28岁男性玩家的原因。这也是运营商计费系统在南亚等新兴市场受到热捧的原因，并且能够解释真钱博彩游戏公司兴起，以及《Candy Crush Saga》等游戏推行跨平台战略的原因。

对比游戏1和ARPDAU相对较高的游戏6，就不难发现这两者的收益差别恰恰验证了我的结论。

总体上看，游戏开发者很显然应该以制作出具有较长玩家生命周期，高留存率，强病毒性和高ARPDAU的游戏为目标。

这里的关键在于要认识到，我们无从保证获得理想的玩家生命周期、留存率、病毒性和ARPDAU。游戏公司最多能做的就是根据以下做法，制作出具有理想参数的游戏：

*快速创建原型和玩法测试：这可以在正式开工之前，快速衡量游戏的潜在留存率，由此判断出有些游戏设计是否值得开发者投入精力。

*广泛的A/B测试：在游戏生命周期中进行健康、广泛的A/B测试十分必要，因为即使是分析中的一点差池也可能对利润造成重大影响。

*频频更新的管道：若要延长平均玩家生命周期，那就必须拥有传递经常更新内容的可靠管道。游戏公司投入开发一款游戏时，要始终如一地将游戏视为半成品。

大型游戏公司已经在开发过程中遵循了上述原理，而小型工作室却通常不走寻常路。这是一种糟糕做法，因为这无助于小型工作室迎头赶上，只会随着时间发展愈发强化大型公司与小型团队的差距。

随着手机游戏市场的发展，当今的游戏开发者很有必要根据数据分析原理来规划自己的整个开发过程，其职责是创造出一种数字娱乐产品，令其最大化地虏获极具病毒性和活跃性的用户。（本文为游戏邦/gamerboom.com编译，拒绝任何不保留版权的转载，如需转载请联系：游戏邦）

The Philosophy Of Game Development By The Numbers

Hassan Baig

Mobile gaming is a huge worldwide opportunity at the moment, having clocked in at $9 billion in 2012, and it is poised to grow further in the coming years. With the world’s 1 billion smartphones scheduled to almost double in number by 2015 and games responsible for a whopping 66 percent of all app revenue, it’s easy for anyone to do the math and see where this is going.

Game development continues to have a bright future, but only for those who can develop profitable titles. Pursuing such profitability is an exact science now, with monitoring analytics and continuous A/B testing having become the staple of game development. In fact, Zynga – the gaming company to have popularized (if not introduced) the use of analytics – has been often categorized as a big data company.

One can imagine metrics to be ‘levers’ that a game developer can push or pull to create a desired outcome. Some levers have a generous range of motion, while others are more limited. In the end a game developer’s task is essentially to figure out the perfect combination of lever positions that will produce the best financial outcome at the least cost. Notions of creativity, novelty and fun are all confined within the prism of this analytics-centric approach: They have wiggle room as long as they improve analytics. That’s the fundamental philosophy behind modern-day game development.

For those looking for a more visceral understanding of game analytics, I’ve set up a simple mathematical simulation that compares game performances across hypothetical retention and viral profiles. It’s in simple spreadsheet format and can be downloaded here. I’ll quickly list out the assumptions governing this simulation, after which I’ll explain the noteworthy conclusions one can draw from the numbers.

Imagine that a gaming studio has six games under its purview:

Note that Game No. 1 is treated as a benchmark and the remaining games differ from it by no more than one metric. For example, Game No. 2 differs from game No. 1 in terms of average player lifetime (and is similar on all other metrics). I have used an average CPA of $1.3 throughout to calculate the games’ respective advertising spend. Lastly, in case more clarity is needed on the definitions of the terms I’ve used in the bullet points above, explanatory descriptions can be found in one of the tabs on the spreadsheet. Now on to the simulation’s broad conclusions.

1) The greater a game’s average player lifetime, the higher its DAU count. And since the DAU is an approximate measure of player engagement which, in turn, is directly correlated to revenue generation, average player lifetime turns out to have an obvious effect on a game’s money-making potential.

In the tables of game No. 1 and game No. 2 in the tab titled “Comparative Revenue” in the spreadsheet notice how game No. 2′s higher average player lifetime gives it superior DAU and revenue numbers in comparison to game No. 1.

2) The greater a game’s d2 retention, the higher its DAU count. And as explained earlier, DAU is directly correlated to revenue generation. Hence it can be surmised that d2 retention has a very obvious effect on a game’s money-making potential. It’s for this reason that most gaming companies utilize A/B testing to optimize their games’ retention rates early in the launch cycle. Also, given d2 retention usually doesn’t optimize beyond single-digit percentages, games with low retention rates are culled very quickly. Look at the comparison between games No. 1 and 3 in the spreadsheet: The latter’s higher d2 retention gives it a better DAU profile, which in turn translates to more revenue overall.

3) Big advertising budgets do not improve a game’s profitability. That is, if a game is a poor financial performer over a certain demographic of players, buying more users for it from the same demographic will not help the bottom line. It’s the reason gaming companies optimize a game’s metrics before buying expensive eyeballs for it, and it’s also the reason certain games get killed way before they’ve seen a full-fledged launch.

Those interested can check out the illustrative comparison between game No. 1 and No. 4 in the “Comparative Revenue” tab in my spread sheet.

4) The greater the virality of a game, the greater its profitability. That is, greater virality ensures more freely acquired users, hence minimizing a key cost consideration: cost per user acquisition. A somewhat similar effect can be garnered via having a captive player network which can be cross-promoted at negligible cost to another game – just that in the former case, virality causes the overall player network to itself expand as well.

Overall, the ability to get free users is extremely important for any gaming company’s financial health, so it’s no wonder that Mark Pincus stressed investing and leveraging Zynga’s player network as a cornerstone of the company’s future strategy in his recent earnings call.

As previously noted, avid number crunchers can have a quick look at the comparison between game # 1 and game # 5 in the “Comparative Revenue” tab in my spreadsheet and appreciate the marked difference between the two games’ eCPA as a result of differing K factors.

5) Higher monetization per user leads to greater profitability. This is quite a straightforward result, but its implications are far-reaching. It’s the reason gaming companies contend for long/multiple sessions and flock around the 43 year old housewife or the 28 year old male gamer, it’s the reason carrier billing is beinghailed as a boon for emerging markets like South Asia, it’s why real-money online gambling is heating up and even why Candy Crush Saga went cross-platform.

Analyze the comparison between game No. 1 and the relatively higher ARPDAU game No. 6. The difference in total revenue between these games illustrates my point.

This concludes the results of my spreadsheet simulation. Many of these results are confessedly intuitive and though looking at my simulated numbers may give a more visceral understanding of fundamental game analytics, it’s only reinforcing what many already know. After all, it’s quite obvious that a game developer should strive for producing a title with lengthy average player lifetimes, high retention rates, great virality and high ARPDAUs.

So other than confirming the obvious, the crux of this exercise is to realize that nothing actually guarantees the achievement of ideal average player lifetimes, retention rates, virality and ARPDAUs. The best a gaming company can really do is set up internal processes and pipelines, such as the ones below, that give it the best shot at producing a game with ideal metrics:

■Rapid prototyping and play testing: This is critical for quickly gauging the potential retention of a proposed game design before full-fledged work is to start on it. Many game designs are just not worth the effort of taking to fruition.

■Extensive A/B testing: Robust, extensive A/B testing throughout the life cycle of a game is very important because even minor bumps in analytics have a directly measurable effect on profitability.

■Pipeline for frequent updates: A reliable pipeline to deliver frequent content updates is a must-have in the bid to prolong average player lifetimes. Once a gaming company commits to a game, it needs to consistently perceive the game as a work-in-progress.

Big-name gaming companies are already following the aforementioned fundamental tenets in their production pipeline – it’s more often the smaller studios which persist with informal methodologies. That’s bad practice because instead of facilitating the smaller studios to catch up, it exacerbates the gap between the big and small fish over time.

As the mobile gaming market continues to spew riches for the foreseeable future, it is imperative that modern day game developers structure their entire operations around the fundamentals of data analytics instead of trying to fit a metrics-based veneer over introverted, blind game development. Their jobs are basically to create digital entertainment products that activate the maximum possible number of highly viral users on a daily basis for the longest sessions.（source：techcrunch）